ChatGPT and other AI bots made huge errors before Scottish election, study finds

Editor's take

AI chatbots like ChatGPT generated demonstrably false information concerning candidates, policies, and electoral timelines in the lead-up to the Scottish election. This inaccuracy is significant as it directly impacts voter comprehension and the integrity of democratic processes during a critical period of public engagement. Such errors highlight the persistent challenge of factual reliability in generative AI, particularly when deployed in sensitive, real-world contexts where misinformation can have tangible consequences.

The Electoral Commission's call for new controls underscores the growing regulatory scrutiny on AI's societal impact. The findings from Demos raise immediate questions about the responsibility of AI developers and platforms to mitigate the spread of disinformation, especially during elections. Future developments will likely focus on the implementation of robust fact-checking mechanisms within these models and the establishment of clear guidelines for their use in political discourse.